基于可能性聚类和卷积神经网络的道路交通标识识别算法
|
摘要
为解决图像采集中噪声和复杂背景对图片的影响以及深度神经网络的高耗时问题,基于可能性聚类算法与卷积神经网络,提出一种道路交通标识识别算法.该方法运用了图像分割技术,并结合卷积神经网络模型对道路交通标识进行更准确的识别.首先,通过色彩增强、图像分割、特征提取、数据增强和归一化等批量预处理操作,形成一个完整的数据集;然后,结合Squeeze-and-Excitation思想和残差网络结构,充分训练出MRESE(My Residual-Squeeze and Excitation)卷积神经网络模型;最后,将优化的网络模型用于道路交通标志的识别.实验结果表明,该方法使训练时间缩短了5%左右,识别精度可达99.02%.
In order to solve the influence of noise and complex background on image acquisition and the problem of high time consuming when we use the deep learning neural network,this paper proposes a road traffic identification algorithm combining probability clustering and deep learning neural network. The proposed method is for more accurate identification of road traffic signs using not only image segmentation technology but also the convolution neural network model. Firstly,a complete data set consists by batch preprocessing operations such as color enhancement,image segmentation,feature extraction,data enhancement and normalization. Then,we sufficiently trained our MRESE(My Residual-Squeeze and Excitation)convolution neural network model which combined the ideas of squeeze-and-excitation and residual network structure. Finally,the optimized network model is used to identify the road traffic signs. The experimental results show that the proposed method reduces the training time by about 5%,and the recognition accuracy can be reached 99.02%.
In order to solve the influence of noise and complex background on image acquisition and the problem of high time consuming when we use the deep learning neural network,this paper proposes a road traffic identification algorithm combining probability clustering and deep learning neural network. The proposed method is for more accurate identification of road traffic signs using not only image segmentation technology but also the convolution neural network model. Firstly,a complete data set consists by batch preprocessing operations such as color enhancement,image segmentation,feature extraction,data enhancement and normalization. Then,we sufficiently trained our MRESE(My Residual-Squeeze and Excitation)convolution neural network model which combined the ideas of squeeze-and-excitation and residual network structure. Finally,the optimized network model is used to identify the road traffic signs. The experimental results show that the proposed method reduces the training time by about 5%,and the recognition accuracy can be reached 99.02%.
引文
[1] Paulo C F,Correia P L. Traffic sign recognition based on pictogram contours ∥ 2008 9th International Workshop on Image Analysis for Multimedia Interactive Services. Klagenfurt,Austria:IEEE,2008:67-70.
[2] Yang H M,Liu C L,Liu K H,et al. Traffic sign recognition in disturbing environments ∥ Zhong N,Ra?Z W,Tsumoto S,et al. Foundations of Intelligent Systems. Lecture Notes in Computer Science,Springer Berlin Heidelberg,2003,2871:252-261.
[3] Bartneck N,Ritter W. Colour segmentation with polynomial classification ∥ 11th IAPR Interna-tional Conference on Pattern Recognition. Vol. II. Conference B:Pattern Recognition Methodology and Systems. The Hague,Netherlands:IEEE,1992:635-638.
[4] Sandoval H,Hattori T,Kitagawa S,et al. Angle-dependent edge detection for traffic signs recognition ∥ Proceedings of the IEEE Intelligent Vehicles Symposium 2000. Dearborn,MI,USA:IEEE,2000:308-313.
[5] Barnes N,Zelinsky A,Fletcher LS. Real-time speed sign detection using the radial symmetry detector. IEEE Transactions onIntelligent Transportation Systems,2008,9(2):322-332.
[6] Gao X W,Podladchikova L,Shaposhnikov D,et al. Recognition of traffic signs based on their colour and shape features extracted using human vision models. Journal of Visual Communication and Image Representation,2006,17(4):675-685.
[7] Shen D G,Wu G R,Suk H I. Deep learning in medical image analysis. Annual Review of Biomedical Engineering,2017,19(1):221-248.
[8] Stallkamp J,Schlipsing M,Salmen J,et al. Man vs. computer:Benchmarking machine learning algorithms for traffic sign recognition. Neural Networks,2012,32:323-332.
[9] 张懿,刘旭,李海峰. 数字RGB与YCbCr颜色空间转换的精度. 江南大学学报(自然科学版),2007,6(2):200-202.(Zhang Y,Liu X,Li H F. The precision of RGB color space convert to YCbCr color space. Journal of Southern Yangtze University(Natural Science Edition),2007,6(2):200-202.)
[10] Piccioli G,De Micheli E,Campani M. A robust method for road sign detection and recognition ∥ Eklundh JO. Computer Vision-ECCV’94. Lecture Notes in Computer Science,Springer Berlin Heidelberg,1994,800:493-500.
[11] Zhou J,Bao X,Li D W,et al. Traffic video image segmentation model based on Bayesian and spatio-temporal Markov random field. Journal of Physics:Conference Series,2017,910(1):Article ID 012041.
[12] 陈力,刘伟峰,杨爱兰. 基于OSPA距离和特征点采样的路标识别算法. 哈尔滨师范大学自然科学学报,2017,33(2):55-57.(Chen L,Liu W F,Yang A L. A recognition algorithm for road signs based on the OSPA metric and characteristic point sampling. Natural Science Journal of Harbin Normal University,2017,33(2):55-57.)
[13] Bengio Y,Delalleau O. On the expressive power of deep architectures ∥ Kivinen J,Szepesvári C,Ukkonen E,et al. Algorithmic Learning Theory. Lecture Notes in Computer Science. Springer Berlin Heidelberg,2011,6925:18-36.
[14] Islam K T,Raj R G. Real-time(vision-based)road sign recognition using an artificial neural network. Sensors,2017,17(4):853.
[15] Hu M K. Visual pattern recognition by moment invariants. IRE Transactions on Information Theory,1962,8(2):179-187.
[16] Caron M,Bojanowski P,Joulin A,et al. Deep clustering for unsupervised learning of visual features. 2018,arXiv:1807. 05520.
[17] Kim D W,Lee K H,Lee D. On cluster validity index for estimation of the optimal number of fuzzy clusters. Pattern Recognition,2004,37(10):2009-2025.
[18] Campello R J G B,Hruschka E R. A fuzzy extension of the silhouette width criterion for cluster analysis. Fuzzy Sets and Systems,2006,157(21):2858-2875.
[19] Forgy E. Cluster analysis of multivariate data:Efficiency versus interpretability of classifica-tions. Biometrics,1965,21(3):768-769.
[20] Krishnapuram R,Keller J M. The possibilistic C-means algorithm:insights and recommen-dations. IEEE Transactions on Fuzzy Systems,1996,4(3):385-393.
[21] 孙即祥. 模式识别中的特征提取与计算机视觉不变量. 北京:国防工业出版社,2001,317.
[22] Zeiler M D,Fergus R. Visualizing and understanding convolutional networks ∥ Fleet D,Pajdla T,Schiele B,et al. Computer Vision-ECCV 2014. Lecture Notes in Computer Science. Springer Berlin Heidelberg,2014,8689:818-833.
[23] Krizhevsky A,Sutskever I,Hinton G E. ImageNet classification with deep convolutional neural networks ∥ Proceedings of the 25th International Conference on Neural Information Processing Systems. Lake Tahoe,NV,USA:Curran Associates Inc.,2012:1097-1105.
[24] He K M,Zhang X Y,Ren S Q,et al. Deep residual learning for image recognition ∥ 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas,NV,USA:IEEE,2016:770-778.
[25] Jia Y Q,Shelhamer E,Donahue J,et al. Caffe:Convolutional architecture for fast feature embedding ∥ Proceedings of the 22nd ACM International Conference on Multimedia. Orlando,FL,USA:ACM,2014:675-678.
[26] 赵永科. 深度学习:21天实战Caffe. 北京:电子工业出版社,2016,392.
[2] Yang H M,Liu C L,Liu K H,et al. Traffic sign recognition in disturbing environments ∥ Zhong N,Ra?Z W,Tsumoto S,et al. Foundations of Intelligent Systems. Lecture Notes in Computer Science,Springer Berlin Heidelberg,2003,2871:252-261.
[3] Bartneck N,Ritter W. Colour segmentation with polynomial classification ∥ 11th IAPR Interna-tional Conference on Pattern Recognition. Vol. II. Conference B:Pattern Recognition Methodology and Systems. The Hague,Netherlands:IEEE,1992:635-638.
[4] Sandoval H,Hattori T,Kitagawa S,et al. Angle-dependent edge detection for traffic signs recognition ∥ Proceedings of the IEEE Intelligent Vehicles Symposium 2000. Dearborn,MI,USA:IEEE,2000:308-313.
[5] Barnes N,Zelinsky A,Fletcher LS. Real-time speed sign detection using the radial symmetry detector. IEEE Transactions onIntelligent Transportation Systems,2008,9(2):322-332.
[6] Gao X W,Podladchikova L,Shaposhnikov D,et al. Recognition of traffic signs based on their colour and shape features extracted using human vision models. Journal of Visual Communication and Image Representation,2006,17(4):675-685.
[7] Shen D G,Wu G R,Suk H I. Deep learning in medical image analysis. Annual Review of Biomedical Engineering,2017,19(1):221-248.
[8] Stallkamp J,Schlipsing M,Salmen J,et al. Man vs. computer:Benchmarking machine learning algorithms for traffic sign recognition. Neural Networks,2012,32:323-332.
[9] 张懿,刘旭,李海峰. 数字RGB与YCbCr颜色空间转换的精度. 江南大学学报(自然科学版),2007,6(2):200-202.(Zhang Y,Liu X,Li H F. The precision of RGB color space convert to YCbCr color space. Journal of Southern Yangtze University(Natural Science Edition),2007,6(2):200-202.)
[10] Piccioli G,De Micheli E,Campani M. A robust method for road sign detection and recognition ∥ Eklundh JO. Computer Vision-ECCV’94. Lecture Notes in Computer Science,Springer Berlin Heidelberg,1994,800:493-500.
[11] Zhou J,Bao X,Li D W,et al. Traffic video image segmentation model based on Bayesian and spatio-temporal Markov random field. Journal of Physics:Conference Series,2017,910(1):Article ID 012041.
[12] 陈力,刘伟峰,杨爱兰. 基于OSPA距离和特征点采样的路标识别算法. 哈尔滨师范大学自然科学学报,2017,33(2):55-57.(Chen L,Liu W F,Yang A L. A recognition algorithm for road signs based on the OSPA metric and characteristic point sampling. Natural Science Journal of Harbin Normal University,2017,33(2):55-57.)
[13] Bengio Y,Delalleau O. On the expressive power of deep architectures ∥ Kivinen J,Szepesvári C,Ukkonen E,et al. Algorithmic Learning Theory. Lecture Notes in Computer Science. Springer Berlin Heidelberg,2011,6925:18-36.
[14] Islam K T,Raj R G. Real-time(vision-based)road sign recognition using an artificial neural network. Sensors,2017,17(4):853.
[15] Hu M K. Visual pattern recognition by moment invariants. IRE Transactions on Information Theory,1962,8(2):179-187.
[16] Caron M,Bojanowski P,Joulin A,et al. Deep clustering for unsupervised learning of visual features. 2018,arXiv:1807. 05520.
[17] Kim D W,Lee K H,Lee D. On cluster validity index for estimation of the optimal number of fuzzy clusters. Pattern Recognition,2004,37(10):2009-2025.
[18] Campello R J G B,Hruschka E R. A fuzzy extension of the silhouette width criterion for cluster analysis. Fuzzy Sets and Systems,2006,157(21):2858-2875.
[19] Forgy E. Cluster analysis of multivariate data:Efficiency versus interpretability of classifica-tions. Biometrics,1965,21(3):768-769.
[20] Krishnapuram R,Keller J M. The possibilistic C-means algorithm:insights and recommen-dations. IEEE Transactions on Fuzzy Systems,1996,4(3):385-393.
[21] 孙即祥. 模式识别中的特征提取与计算机视觉不变量. 北京:国防工业出版社,2001,317.
[22] Zeiler M D,Fergus R. Visualizing and understanding convolutional networks ∥ Fleet D,Pajdla T,Schiele B,et al. Computer Vision-ECCV 2014. Lecture Notes in Computer Science. Springer Berlin Heidelberg,2014,8689:818-833.
[23] Krizhevsky A,Sutskever I,Hinton G E. ImageNet classification with deep convolutional neural networks ∥ Proceedings of the 25th International Conference on Neural Information Processing Systems. Lake Tahoe,NV,USA:Curran Associates Inc.,2012:1097-1105.
[24] He K M,Zhang X Y,Ren S Q,et al. Deep residual learning for image recognition ∥ 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas,NV,USA:IEEE,2016:770-778.
[25] Jia Y Q,Shelhamer E,Donahue J,et al. Caffe:Convolutional architecture for fast feature embedding ∥ Proceedings of the 22nd ACM International Conference on Multimedia. Orlando,FL,USA:ACM,2014:675-678.
[26] 赵永科. 深度学习:21天实战Caffe. 北京:电子工业出版社,2016,392.